The Dream of Auto-marking Assessments

I was chatting to a teacher a couple of months ago who was planning to develop the ultimate teaching time-saving hack, using AI to generate and then mark classroom assessment. Sounds far-fetched but the basic approach is to generate assessment, of varying types, using an AI system such as ChatGPT, formatting it as a printable class test and then distributing it to learners. They would then complete the paper-based test in class and these would be digitised, either via scanning or even taking a photograph using a smartphone. These completed, digitised tests could then be fed into a simple AI-system, e.g. using the OpenAI API, streamlining the evaluation process.

I was fascinated by this potential efficiency and spent time looking into it and auto-marking in general. Here are my thoughts! AI-assisted auto-marking seems to present a reliable, objective, and efficient means to evaluate learner submissions. It eliminates the possibilities of human error and bias (obviously note in the original LLM training data), ensuring consistent evaluations. With Natural Language Processing (NLP), AI systems can analyze the structure, coherence, and quality of language-based assignments such as essays. I've seen this work with text submitted via ChatGPT. These systems can also swiftly handle multiple-choice tests and detect plagiarism through pattern recognition.

But beyond efficiency, what truly makes AI indispensable is its impact on the educational ecosystem. For educators, the automation of grading could be liberating. It alleviates the administrative burden, allowing educators to dedicate their time and efforts to personalised teaching, mentoring, and creating engaging learning environments, which has been shown to be the most important indicator in learner success.?

For learners, the immediate feedback provided by AI-driven auto-marking is invaluable. It facilitates real-time learning by allowing learners to understand their performance promptly. Moreover, objective and consistent evaluations ensure fairness in grading.

However, as we all embrace this technology head-first, it’s also critical to recognise and address its limitations. AI, for all its apparent genius!, cannot replicate the human understanding of the nuances and complexities in creative work. Additionally, biases in training data can influence AI algorithms. Ensuring the fairness and accuracy of AI-driven assessments demands continuous algorithm improvement, meticulous training data selection, and perhaps maintaining a parallel human evaluation for validation. Fostering a dialogue between AI experts and educators can lead to more nuanced and ethically-grounded implementations.

Using AI in auto-marking assessments is not just an incremental change; it’s a paradigm shift. It’s about optimising the grading process, empowering educators, and enriching the learning experience for learners. As this technology continues to evolve, it holds the promise of an educational environment that is more efficient and more responsive to the needs of educators and learners alike.

I'm super interested in seeing these sorts of platforms in the wild - please share links to any you may have discovered or used.